Reflections on adapting group model building scripts into online workshops

Abstract

Group model building (GMB) is an important social process in system dynamics (SD) for creating a shared understanding of complex systems and providing a platform for stakeholders to exchange information and ideas (Antunes et al., 2015; Sedlacko et al., 2014). Gathering stakeholders around the table to discuss a contentious issue can provide important insights for SD modeling (Hovmand, 2014; Rouwette and Vennix, 2006; Van den Belt and Dietz, 2004; Vennix, 1996; Vennix et al., 1992). While online meeting platforms and collaborative tools have made great leaps forward in recent years, systems mapping has largely continued with in-person facilitation. And with good reason: in-person facilitation has proven to be an effective way to generate discussions among stakeholders (Stave, 2010). Many in the SD community have long seen the potential in using interactive platforms for GMB with stakeholder participation (e.g. Kenzie et al., 2018). Yet, to our knowledge, there are no documented efforts of a fully online GMB workshop. In our case, the COVID-19 virus and sudden change in travel and work patterns meant that an anticipated GMB workshop could not be executed as planned. Instead of canceling or postponing the workshop, we used this disruption as an opportunity to test the potential of running GMB processes online. One key advantage of online GMB is that it can make room for more—and more diverse stakeholders—at the table, no matter where they may be in the world. This can improve access and participation, although groups with strong power dynamics may require skillful facilitation. Online GMB can also greatly reduce the need for travel, which could reduce the amount of time and money needed for a workshop in addition to reducing carbon footprints. The lack of travel logistics can also mean that workshops could be planned with less advance notice and perhaps be more responsive to current issues. Online GMB, with stakeholders sitting alone in front of their computer screens, is fundamentally different from in-person GMB workshops, and there are limitations to how much one can substitute for the other. We found, however, that we could recreate many of the strengths of an in-person GMB workshop in an online environment. Our positive initial experiences with online GMB indicate that it's worthwhile to build on and further develop methodologies for online stakeholder engagement. We propose that online GMB, a practice with its own set of strengths and weaknesses, should be further developed as its own set of methodologies parallel to traditional (in-person) GMB. This research note includes a description of our experiences and lessons learned as well as areas for further development. The workshop was planned to gather representatives from the business sector, public policy sector, and academia to work on two cases. The first case addressed the Bergen, Norway, goal of becoming “fossil fuel free” by 2030, and the second case focused on innovation across sectors for ocean technology and offshore industry. The workshop was the first phase in a larger project. Subsequent phases include a hackathon in which SD students and practitioners use the causal loop diagrams from the workshop to generate models and ideas for policy interventions. In phase three of the project, winning submissions from the hackathon are refined and presented at an “innovation festival” in Bergen. The workshop had been planned for early June 2020 as an in-person, five-hour GMB session with a focus on systems mapping. When it became clear that the workshop could not be held as planned due to Covid-19 and social-distancing requirements, we assembled an international team to redesign the session as an online workshop on the same date. Participants had little or no previous experience with systems mapping, systems dynamics, or systems thinking. The workshop was advertised as a “Crash Course in Systems Thinking,” and all participants referred to the possibility of learning a new skill or perspective (as opposed to having strong stakes in the issue we were mapping) as a major reason for registering. Our participants were fairly homogeneous in education levels and technological savvy. Further, all participants were from Norway, a society typically characterized as egalitarian and “flat,” with an expectation that everyone's contribution should be heard. Working with more diverse (ethnicity, familiarity with technology, language abilities, etc.) groups, or groups with more complex power dynamics, in an online environment will require more sophisticated facilitation techniques, but that is not the focus of this research note. A significant consideration for our design was to minimize the transitions between workspaces and the learning curve for participants. While existing SD softwares offer increasingly sophisticated causal-loop diagram abilities, they do not allow for real-time group editing. Some members of the group had experience with other online tools such as Kumu, Sheetless, Padlet, Loopy, Mental Modeler, and Meetingsphere, but none of those platforms met all our needs for divergent and convergent activities. After much searching, we realized that there was no perfect online platform, but that one platform, Miro (miro.com), came close. Miro is a very flexible, collaborative, and interactive online platform that includes a large board, onto which different types of notes, discussions, and diagrams can be placed. Everyone with access to the board can add or take away text or drawings. Other valuable features include a timer and a “share screen” feature that allows a facilitator to bring everyone to the same part of the board. Miro includes a video chat function, though it only shows the videos of up to three people at a time, and the facilitator could not mute all participants at once. Lastly, although Miro's tools include “sticky” arrows, we could not easily assign +/− symbols to the arrows (though arrow color could be changed to show polarity). Another important limitation of Miro is that it does not offer the possibility for turning the systems map into a simulatable model in later stages of the GMB process. As Miro could not meet all our needs, it was used in conjunction with Zoom, email, Google Drive, and WhatsApp during various phases of workshop planning and execution (Figure 1). Both divergent and convergent group activities were selected from standard scripts, which define essential elements of the GMB workshop, the steps needed to complete the script, and the outputs produced from the scripts (Andersen and Richardson, 1997; Hovmand et al., 2011; Hovmand et al., 2012). The modeling team then adapted the scripts to the online workshops and developed the workspace. Every aspect of the workspace layout had to be built “by hand,” as Miro does not include empty graphs or other templates that we needed. The main design challenge was to find ways the platform could support the strengths of the different activities. For divergent activities, we needed to give control to the participants to produce their ideas. For this, we designed individual workspaces that the participants could zoom in to and work without disruption or distraction. For convergent activities, a balance was necessary between viewing large clustering or mapping spaces with the minimum possible loss of information on individual variables. Text boxes that automatically resize depending on the zoom were particularly useful in this respect, and a more “tight” design was selected to reduce information loss. All members in the modeling team were trained in GMB methods, and the entire team needed to be familiarized with the online workspace and their roles to generate the best results. The roles included meeting opener/closer, facilitators, modelers, and recorders, as indicated in the scripts. We also included an additional role specific to the online architecture: the stage manager who controlled the view of the workspace areas. The stage manager used the “Share Screen” mode and the “Bring-To-Self” feature to guide stakeholders to different parts of the board while activities were described and demonstrated with examples. For individual activities, the “Share Screen” mode was turned off, and participants could navigate and zoom in and out of the board as they desired. The differences between the original and adapted scripts for the virtual environment, were (i) the materials, which were replaced by the tools in Miro and (ii) some parallel activities in Miro that needed to be paired with the corresponding step in the script. We created a Google document with the steps in each script, timing, and the specific activities that the stage manager and modeler needed to follow in Miro along with the facilitator's activities. This adapted script was applied in both case studies. As an illustration, we present an example of how we adapted the Graph Over Time script to Miro's workspace. For practical reasons, we combined the Graphs Over Time script with the Variable Elicitation script. Participants first generated a set of graphs with the key factors they considered were causing the problem, and we used those graphs as a starting point to elaborate the CLD. Participants could then add variables using a text box as they were building the diagram with the modeler's help. To develop the layout and material for the script in Miro, the design team focused on what usually “works” in the in-person execution of the script and attempted to mimic, to the extent possible, the experience of a GMB workshop. Due to the small size of the group, we were able to create individual virtual desks for independent work. Graphs Over Time templates were placed on each workspace to reduce the need for moving elements in the platform, and a “stack” of additional templates was placed next to each desk. Following the script, a board for clustering ideas and a large area for systems mapping were developed, accompanied by “parking lot” and “duplicates” areas (Figure 2). In Table 1, we detail the activities and roles for online implementation of the Graphs Over Time script. After the Miro layout and the script were adapted, the entire team conducted two practice rounds to test technical issues, visibility and ease of use of elements on the board, and timing for different steps in the process. Through the trials, the entire team could test the workshop experience in real time and practice using tools and navigating the board as a participant. In addition, team members developed familiarity with the script, roles, and tasks and practiced interactions among each other to ensure a smooth experience for participants. These practice rounds resulted in a number of ideas and adjustments for streamlining the workshop and improving user experience. The day before the workshop, participants were sent an email with instructions for creating an account on Miro, a brief video (made by a team member) showing participants how to use some of the tools on Miro, and links to a Zoom meeting and the Miro board. The participants also received a workshop agenda describing each activity (see Table 1). The workshop started in Zoom, a platform we knew most stakeholders were familiar with (Figure 1). Here, we introduced the workshop and the GMB process. We then asked participants to join their relevant board in Miro via a link that had previously been sent to them in an email. A facilitator stayed on Zoom to chat with participants who had trouble accessing the Miro board and joining the video chat there. The first activity on Miro was an icebreaker in which participants drew a picture about their interests and wrote a few words about themselves. In addition to getting people familiar with each other, this had the advantage of encouraging participants to use the writing and drawing tools in Miro that they would be using in the workshop activities (Figure 3). The modeling team then followed the adapted scripts according to their roles. A summary of the activities of the workshop is presented in Table 2. Participants introduce themselves using Miro features while familiarizing themselves with the Miro's tools. The workshop team used communications apps (Zoom and WhatsApp) to communicate privately both during active parts of the workshop and during breaks (Figure 1). Direct messaging via WhatsApp allowed us to discuss issues related to the workshop without disturbing participants and give each other feedback on what we were observing. During breaks, we used Zoom to discuss and edit the CLD. By muting ourselves in Miro, we could discuss freely and collaboratively without disturbing participants who kept their sound on. Facilitators, modelers, and stage managers accessed the prepared scripts via Google Drive, and the same service was used by recorders to document the session. As members of the workshop team were using Miro, Google Drive, and a separate communications app at the same time, we found having two screens very helpful. Planning and conducting an online GMB workshop provided a rich learning experience for facilitators and modelers in addition to participants. In a debriefing after the workshop, we collated a list of insights that we would take forward into future online workshops. While some of these lessons learned are general “best practices” for any stakeholder engagement exercise, we highlight them here because we found them to be especially important in an online environment. We see several facets of online GMB that should be further explored and developed. Our experience points to the importance of good facilitation techniques. GMB likely works best with stakeholder groups with low levels of power differences (Vennix, 1996). It would be especially valuable to explicitly test to what degree facilitation techniques could address power dynamics in an online workshop. We also see the benefit in creating a guide or training to learn facilitation techniques for GMB in online environments. Further research could also determine optimum group sizes for this methodology. Small groups allow for easier online dialog, yet larger groups can provide a greater diversity of perspectives and contexts. This balance point will depend on the chat and video capabilities of the online platforms in addition to the usual considerations for a workshop (e.g. topic complexity, group heterogeneity). Our workshop used common scripts (such as Graphs Over Time and Action Ideas), and we would welcome further explorations of implementing other GMB scripts (e.g. Causal Mapping with Seed Structure) virtually or even asynchronously. Evaluations of long-term mental model changes, especially as compared to in-person workshops, would improve our understanding of the impact of online systems mapping on stakeholders. Surveys before and several weeks after the workshop could assess levels of engagement and information retention (Scott et al., 2013). Paired studies, with some stakeholders assigned to an online workshop and others assigned to an in-person workshop on the same topic, could aid this understanding. If conditions in a multisession workshop allow, approaches that combine initial in-person meetings with subsequent online meetings, or that combine in-person convergent activities with online divergent activities, could be investigated. Due to power dynamics and facilitation challenges, we do not recommend having some participants in person and others online in the same workshop session. As more examples of online GMB workshops are assembled, we'd like to see more systematization and standardization of approaches in order to ensure comparability across workshops (McCardle et al., 2009; Scott et al., 2016). Moving GMB workshops online is not without challenges, but we feel that the significant time and effort invested in bringing a system's mapping experience into an online environment was worth it. Our initial instinct was confirmed: online GMB works! We received positive feedback from a number of participants, including “I came out of [the workshop] with very useful experience and ideas for how we can carry out workshops and think holistically about our customers' problems” and “I learned a lot about systems thinking and system dynamics and want to learn even more now.” This relatively simple workshop allowed us to test the possibilities we saw in online GMB and develop ideas for further exploration. Choosing between online or in-person GMB depends on workshop aims. In-person experiences provide a wealth of interactions that open the space for creating a rich understanding of the issue at hand. That same depth of interaction cannot be recreated online, at least not with today's technology. Online GMB can, however, open up for a larger breadth of interactions by including people who, because of time, finances, or distance would otherwise be excluded. As online GMB takes place in a “neutral” space, it may also make it easier to bring people together from across organizations and institutions. In short, online GMB offers significant advantages that are fundamentally different from the advantages of in-person GMB. We believe that further explorations of how best to include this breadth in a digital platform can enrich the practice of GMB as a whole. We view online group model building as a parallel methodology that warrants further development, and we look forward to learning from each other's' experiences in the system dynamics community as this methodology advances. The authors would like to thank the city of Bergen for providing support and funding and Sanjeet Jaiswar for assisting in the workshop. We thank the editors of the System Dynamics Review for their comments on an earlier version of this note.